QIAGEN’s bioinformatics team had a great time at the annual meeting of the American College of Medical Genetics and Genomics in North Carolina earlier this month. No doubt about it, this is the best conference of the year for variant interpretation! We thoroughly enjoyed attending sessions and talking to our colleagues and peers about a topic that’s important to us all. As we did, several key themes emerged.
Expert involvement. We’ve been immersed in bioinformatics since the ’90s and can say with certainty that there has always been a tension between automating processes and incorporating human judgment and expertise. That debate was on display at ACMG this year, with various speakers demonstrating how they balance automation with expert review. In one session, Harvard’s Heidi Rehm spoke about the disease-specific expert panels that are now reviewing data being fed into ClinVar. Today, that database has more than 9,300 variants that have been reviewed by these panels, but Rehm noted there’s a long way to go since there are hundreds of thousands of variants in ClinVar. Still, she said, this approach is one way ClinVar data is being cleaned up over time.
Need to scale. Without question, variant interpretation pipelines have to become more scalable than they are today. This was widely acknowledged at ACMG, where speakers shared their experiences developing or using a host of different methods for automating some or all of the interpretation process. As the number of variants reported rises, a trend that will grow exponentially in the coming years as more testing shifts to whole-genome sequencing, this problem is only going to become more pronounced.
Importance of the literature. Sure, we all know it’s important to consult the scientific literature for variant interpretation — but speakers at ACMG really underscored the need to consult the broader body of literature, not just the obvious papers. Just as limiting the search for variants to an exome necessarily narrows the possible findings from what could be discovered by searching everything, so too does making assumptions about which knowledge is needed to interpret a variant. There was an obvious push at this year’s meeting to expand interpretation processes as much as possible to incorporate more information for stronger results.
Reinterpretation. Several presentations cited data showing that sequenced exomes or genomes that provided no diagnostic results for a patient’s case could be reexamined as soon as a year later with more useful results. It’s a sign of how quickly our understanding of genes, variants, and diseases is expanding. In her presentation, Columbia’s Wendy Chung supported the idea of subscription services that would continually re-interpret genetic or genomic data to improve the diagnostic yield over time.
These are all topics we spend a lot of brain power on. We believe that variant interpretation must be automated — there’s simply no feasible alternative given the demand we’ll be facing in the coming years — and that the successful approaches will manage to incorporate expert curation and judgment as a key part of the process. For example, our QIAGEN Clinical Insight tool was designed as a clinical decision support tool; it offers the recommended interpretation, but allows users to review every bit of evidence supporting that conclusion and to adjust filters or change assumptions based on their own expertise. The tool is built on our one-of-a-kind QIAGEN Knowledge Base, an expert-curated representation of the scientific literature alongside publicly available databases for the broadest information foundation.
We’d like to thank all of the ACMG attendees who stopped by the QIAGEN booth and shared their work with us. It was a pleasure to meet you all!